[Tagging] simple_brunnel : one node bridge like xing highway over waterway

Simone Saviolo simone.saviolo at gmail.com
Fri Apr 4 09:51:54 UTC 2014

2014-04-03 22:42 GMT+02:00 Richard Z. <ricoz.osm at gmail.com>:

> > Don't dismiss that argument so casually. The current rule is that the
> > way below the bridge should not share a node with the bridge itself.
> the current idea that culverts float bellow roads without having anything
> common with them is not correct in most cases. These culverts are part of
> an integral highway-culvert-waterway construction. The same is true for
> most bridges, only a small fraction does float independently above valeys
> but most are connected with the lower way by the actual bridge
> construction.

The bridge structure may also be related to the riverbed structure, but
ways are not. As you drive on the road on the bridge you have no idea
whether down below there's a river or a stream or a valley.

> > I could imagine adding an exception to that rule if it were hard to
> > avoid a shared node. But in this case, it can very easily be avoided by
> > mapping the bridge in the same manner two million other bridges have
> > already been added: as a way.
> easily? So you have biked 60 miles along a forest track and know reliably
> that there was not a single ford on your route today. You look at OSM data
> in the evening and see there are 120 streams which you crossed with missing
> bridges/culverts.
> What do you do? Leave those 120 crossings in incomplete state even though
> someone might be really interested to know whether there are some fords on
> the way?
> Add fictional bridges or culverts?
> Say "ford=no"?

This is nonsense. If two ways don't cross, they don't cross. Missing the
bridge/culvert tag is a minor error: it just leaves you without information
as to how that road and that waterway intersect their paths. However, if
they share no common node at the intersection, you can assume that there's
no way you could stop driving and dive into the water. In case there's a
ford instead, map it: put a node on the intersection and use ford=yes, so
people will know that *those* two ways cross with a ford.

Missing the tag is missing information with a fallback that makes sense
(you'd notice an unmapped ford with your eyes and go fill it in). Putting
in a node is *wrong* information, and consumers would assume that the ways
cross, thus ending up with a wrong routing graph (maybe they'll penalize
the route thinking there's a ford).

> The other point - even if you know it is a bridge or culvert - is it worth
> "painting" an insignificant structure which is perhaps 3m in size when the
> GPS error is more likely 10 meters? In a deep valley and forest in the
> mountains
> you are often lucky to get GPS precision better than 60m.

Stop saying GPS. Forget even about aerial imagery. When I had no aerial
imagery in my area, I either did not draw such features (leaving them for
future improvements), or approximate. "The road there is about 6 meters
wide, so I'll draw two nodes about 6 meters apart, split the waterway there
and tag the middle piece as a culvert". It's not that hard, it's not that
much imprecise, sure it may be improved with better measurements, but it is
not wrong, especially it is not topologically wrong.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20140404/b7ec5962/attachment.html>

More information about the Tagging mailing list